diff --git a/tests/infrastructure/epc_client/test_sap_accuracy_corpus.py b/tests/infrastructure/epc_client/test_sap_accuracy_corpus.py
index cc57d28d..c489e1ef 100644
--- a/tests/infrastructure/epc_client/test_sap_accuracy_corpus.py
+++ b/tests/infrastructure/epc_client/test_sap_accuracy_corpus.py
@@ -248,7 +248,16 @@ _MIN_WITHIN_HALF_SAP = 0.74
 # an identical dwelling rates SAP 87 with "Connected to Dwelling = Yes" (credit
 # -£167) vs SAP 74 with "No" (credit £0). Enum decoded empirically: 0 = no PV,
 # 1 = not connected, 2 = connected (the gov-API does not expose it elsewhere).
-_MAX_SAP_MAE = 0.740
+# Then 0.740 -> 0.726 via the heat-pump PSR-extension fix (SAP 10.2 Appendix N2,
+# PDF p.101 footnotes 44/45): an air/ground/water source heat pump whose plant
+# size ratio exceeds the PCDB record's largest PSR is no longer clamped to the
+# top-of-table COP — its efficiency is reciprocal-interpolated toward 100% at
+# twice the largest PSR (and 100% below the smallest PSR). Accredited Elmhurst
+# worksheet for cert 100110101713 (golden fixture case 56, PCDB 100061, PSR
+# 3.107 over largest 2.0): (206) 334.4% -> 139.66% = Elmhurst exact. Only two
+# certs move (both oversized-PSR heat pumps): 100110101713 +18.32 -> -4.97 and
+# 4510053280 -0.61; within-0.5 holds at 74.1%.
+_MAX_SAP_MAE = 0.726
 _MAX_CO2_MAE_TONNES = 0.09      # t CO2 / yr vs co2_emissions_current
 _MAX_PE_PER_M2_MAE = 3.5        # kWh / m2 / yr vs energy_consumption_current